Negativity Bias, Negativity Dominance, and Contagion
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Personality and Social Psychology Review Copyright © 2001 by
2001, Vol. 5, No. 4, 296–320 Lawrence Erlbaum Associates, Inc.
Negativity Bias, Negativity Dominance, and Contagion
Paul Rozin and Edward B. Royzman
Department of Psychology and Solomon Asch Center for Study of Ethnopolitical Conflict
University of Pennsylvania
We hypothesize that there is a general bias, based on both innate predispositions and
experience, in animals and humans, to give greater weight to negative entities (e.g.,
events, objects, personal traits). This is manifested in 4 ways: (a) negative potency
(negative entities are stronger than the equivalent positive entities), (b) steeper nega-
tive gradients (the negativity of negative events grows more rapidly with approach to
them in space or time than does the positivity of positive events, (c) negativity domi-
nance (combinations of negative and positive entities yield evaluations that are more
negative than the algebraic sum of individual subjective valences would predict), and
(d) negative differentiation (negative entities are more varied, yield more complex
conceptual representations, and engage a wider response repertoire). We review evi-
dence for this taxonomy, with emphasis on negativity dominance, including literary,
historical, religious, and cultural sources, as well as the psychological literatures on
learning, attention, impression formation, contagion, moral judgment, development,
and memory. We then consider a variety of theoretical accounts for negativity bias. We
suggest that 1 feature of negative events that make them dominant is that negative enti-
ties are more contagious than positive entities.
Brief contact with a cockroach will usually render taminated—that is, lowered in social status—by
a delicious meal inedible. The inverse phenome- contact with members of lower castes. The contami-
non—rendering a pile of cockroaches on a platter ed- nation often occurs by eating food prepared by a
ible by contact with one’s favorite food—is unheard lower caste. On the other hand, when people of lower
of. More modestly, consider a dish of a food that you castes consume foods prepared by higher castes,
are inclined to dislike: lima beans, fish, or whatever. there is no corresponding elevation in their status.
What could you touch to that food to make it desir- Stevenson (1954) summarized this feature of the
able to eat—that is, what is the anticockroach? Noth- caste system with the phrase “pollution always over-
ing! And the cockroach is far from unique: there is a comes purity” (p. 50). The caste system pulls down-
wide variety of animals (e.g., worms, caterpillars, ward; it is easy to pollute and hard to purify.
slugs, spiders) that share the cockroach potency, Similarly, in many Western and non-Western reli-
along with a variety of microbially or toxin-contami- gious traditions, becoming possessed by a malevolent
nated objects. One of the best generic descriptions of demonic force is a relatively brief and easy affair
this relative power of negative contamination is em- (Oesterreich, 1974), whereas the reversal of the pos-
bedded in an age-old Russian adage: “A spoonful of session requires the painstaking, prolonged, and often
tar can spoil a barrel of honey, but a spoonful of injurious ritual of exorcism. On the other hand, in
honey does nothing for a barrel of tar.” This apparent these same traditions, becoming “holy” or “saintly”
dominance of negative over positive contamination is usually involves a long moral trajectory of positive
played out on a vast human–social scale among the deeds, a state that can be compromised easily by one
large traditional segment of 800,000,000 living or a few immoral acts. The general principle that, for
Hindu Indians. People of higher castes are easily con- forgiveness to be achieved, the degree of acceptable
expiation must dramatically exceed that of the initial
Preparation of this paper was supported by National Institute of Drug fault, is dramatically played out in the Christian con-
Abuse Grant R21–DA10858–0, the Edmund J. and Louise W. Kahn cept of redemption from original sin. In a number of
Chair at the University of Pennsylvania, and the Solomon Asch Center
traditions, there are some losses to one’s purity, such
for Study of Ethnopolitical Conflict at the University of Pennsylvania.
Requests for reprints should be sent to Paul Rozin, Department of as the sullying of female sexual honor through pre-
Psychology, University of Pennsylvania, 3815 Walnut Street, Phila- marital sex or illicit affairs, which cannot be reme-
delphia, PA 19104–6196. E-mail: rozin@psych.upenn.edu died at all.
296NEGATIVITY BIAS
The dominance of negative contamination affords a documented the wide range of positive biases. These
particularly striking demonstration of what we take to appear in higher frequency of positive words, positive
be a very general principle, a principle that holds across experiences, and positive views of the world, and in
a wide range of domains, and in nonhuman as well as hu- other domains. This puts us in the peculiar position of
man animals. The principle, which we call negativity describing what we believe to be a basic tendency in the
bias, is that in most situations, negative events are more face of documented evidence for the opposite tendency,
salient, potent, dominant in combinations, and gener- as well.
ally efficacious than positive events. (There are excep- Guido Peeters and his colleagues (e.g., Lewick et al.,
tions to this claim, but they constitute a minority of cases 1992; Peeters, 1971, 1989; Peeters & Czapinski, 1990)
and often involve special circumstances). directly addressed this apparent contradiction, which
The principle of negativity bias has not escaped the they described as positive–negative asymmetry. They
attention of thinkers in many disciplines. The principle treated the evidence for both positive and negative bi-
has been noted by a number of prominent humanists, in- ases in a sophisticated and balanced way. They noted the
cluding Shakespeare, Pushkin, and Schopenhauer (see interesting fact that, because negative events are much
quotes later in this article). Contemporary psychologists rarer than positive events, it is adaptive to assume the
have pointed to forms of negativity bias in particular do- positive (the most likely occurrence) while being
mains, and in general. These contributions will be dis- watchful for the dangerous negative. Thus, many exam-
cussed in more detail later, but to acknowledge our ples of positive bias result from the same basic fact about
predecessors, we note here that, in a political science the world, the dominance of positive experiences, as
journal, Jordan (1965), summarizing results from dif- does negativity bias. Peeters and his colleagues have set
ferent studies, concluded that “a positive attitude or pos- the stage, appropriately, for a careful examination of the
itive affect does not have an effect on measured contexts in which each type of bias appears, and attempt
behavior oppositely equivalent to the effect of a nega- to bring both phenomena under one conceptual um-
tive attitude or negative affect” (p. 315). Kanouse and brella. They have pinpointed the issue in noting the
Hanson (1972) identified a negative bias effect with re- greater frequency, but lesser “urgency,” of positive
spect to a well-defined range of phenomena. Guido events. Organisms must be most efficient at dealing
Peeters and his colleagues have produced many demon- with the most frequent occurrences, but also the most
strations of negative bias, particularly in the context of important occurrences. Cacioppo and his colleagues
attitude and impression formation (Lewick, Czapinski, (Cacioppo & Bernston, 1994; Cacioppo, Gardner, &
& Peeters, 1992; Peeters, 1971, 1989; Peeters & Bernston, 1997, 1999) also recognized positive and
Czapinski, 1990). negative bias; they focused on a negativity bias as a
Negativity bias has been the focus of attention in a more rapid recruitment of negativity than positivity
few other lines of thought in psychology. The greater with increasing strength of elicitors and a positivity off-
general potency of negative events is at the core of pros- set defined in terms of a bias to treat relatively neutral
pect theory, as described in the prospect function and la- entities as weakly positive.
beled as loss aversion (Kahneman & Tversky, 1979; Our contribution in this article to the prior work is
Tversky & Kahneman, 1991). Taylor (1991) focused on fivefold.
the negative–positive asymmetry, in many of its mani-
festations, in the framework of demonstrating adaptive 1. We extend the range of domains in which
asymmetrical coping processes involved in neutralizing negativity bias has been noted.
the greater negative potency. More recently, in a series 2. We present a taxonomy of negativity bias phe-
of articles, John Cacioppo and his collaborators nomena.
(Cacioppo & Bernston, 1994; Cacioppo, Gardner, & 3. We clearly distinguish the special and, we be-
Bernston, 1997, 1999) noted a negativity bias effect in a lieve, most robust and informative subclass of
number of domains, including three of the four that we negativity bias phenomena: negativity dominance.
document and organize in this article. Finally, inde- 4. In particular, we highlight the previously un-
pendently of our work, and at the same time, a review of noted area of contagion and contamination, which we
articles emphasizing negativity bias, particularly in the believe is the most robust and informative subclass of
social interaction and impressions domain, has been negativity dominance. We argue that contagion and
completed (Baumeister, Bratslavsky, Finkenauer, & contamination matters both as a domain in which
Vohs, in press). negativity bias makes some of its most dramatic ap-
Although negativity bias is often striking, it is far pearances and a possible mechanism that mediates
from universal. On the contrary, there is sufficient evi- negativity bias effects in other spheres.
dence for a positive bias that an entire book, The 5. We review and extend, in light of the first four
Polyanna Principle (Matlin & Stang, 1978), has amply points, the significance, extent, and theoretical ac-
297ROZIN & ROYZMAN
counts of the various negativity bias phenomena and at- Loss aversion has been demonstrated in a number of
tempt to include many of the phenomena of positive domains, but it does not always occur. The demonstration
bias under the same theoretical accounts. of negative potency is unfortunately limited, because it
requires a metric (usually money) to establish the objec-
We regard our work as being largely complemen- tive equality of negative and positive events (e.g., losing
tary to the simultaneous work of Baumeister et al. (in and gaining $100, or 10 lb of weight, or 5° in tempera-
press). The two articles differ both in respect to the ture), or comparison of two contexts for the same mate-
structure of their arguments and the range of examples rial entity (as in the case of the endowment effect).
they proffer in support of the general proposition that
negativity bias is a pervasive and consequential feature
of human existence. However, the main difference be- Greater Steepness of
tween our works lies elsewhere. Bauimeister et al. ana- Negative Gradients
lyzed the phenomenon at hand primarily in terms of
independent, orthogonal influences of bad and good There is minimal but provocative evidence that nega-
things on behavior and cognition. Our work, on the tive events grow more rapidly in negativity as they are
other hand, emphasizes the tendency of the effects of approached in space or time than do positive events.
the negative to dominate (or even utterly overwhelm) There have been only a few empirical attempts to dem-
those of the positive when the two are blocked together onstrate this, but we consider this principle potentially
to form a single configuration. This echoes the Ges- important. The clearest demonstrations and discussions
talt-like claim that, in a wide range of cases, the “cog- of this phenomenon appear in the animal learning litera-
nitive interaction” of two stimuli, such as those ture (Brown, 1948; N. E. Miller, 1944; discussed later).
denoting evaluatively positive or negative personal at- Recently, Cacioppo and colleagues (Cacioppo &
tributes and brought together to form a novel compos- Bernston, 1994; Cacioppo, Gardner, & Bernston, 1997,
ite, cannot be accurately anticipated from prior 1999) discussed this asymmetry and incorporated it into
knowledge of the values of the two stimuli taken apart their computational model of evaluative space. In par-
(Rokeach & Rothman, 1965; Royzman, 2000). ticular, they posited steeper negative gradients (which
they described as negativity bias) along with a tendency
Four Aspects of Negativity Bias for a net positive outcome with very weak negative and
positive inputs (which they described as positivity off-
In this taxonomic section we propose three or four set). In addition to the work of Brown and Miller, they
types of negativity bias: negative potency, greater cited evidence from the impressions literature, suggest-
steepness of negative gradients, negativity dominance, ing that negative events dominate positive events only
and negative differentiation. when both are strong.
It is possible that steeper negative gradients are sim-
ply a manifestation of negative potency because the
Negative Potency steeper gradient follows from the fact that additional
negative units (measured as stimuli) will produce larger
The principle of negative potency asserts that, given psychological effects than additional positive units.
inverse negative and positive events of equal objective However, in light of the phenomenon of positivity off-
magnitude, the negative event is subjectively more po- set, the gradient effect may be distinct from negative po-
tent and of higher salience than its positive counterpart. tency because at low levels negative inputs do not seem
More generally, the claim is that negative events are to be more potent than “equivalent” positive inputs.
more potent with respect to their objective magnitude Negative potency might result from higher subjective
than are positive events. This is described in the pros- levels of negative stimuli at all stimulus levels, so that
pect function and is at the core of the loss aversion phe- the two functions might have the same slope but the neg-
nomenon (Kahneman & Tversky, 1979; Tversky & ative function might have a higher intercept. However,
Kahneman, 1991). For example, in the domain of poli- the gradient results suggest that this is not the case;
tics, Bloom and Price (1975) showed that short-term rather, the negative functions seem to have a lower inter-
economic conditions, when they are downturns, reduce cept but a higher slope.
the vote for the party of the incumbent in American
presidential elections, whereas upturns have virtually no
effect. The endowment effect is perhaps the purest and Negativity Dominance
most robust instantiation of loss aversion (Kahneman,
Knetsch, & Thaler, 1990); people will demand much According to the principle of negativity dominance,
more to give up something they possess (a loss) than the holistic perception and appraisal of integrated neg-
they will pay to obtain the same item (a gain). ative and positive events (or objects, individuals,
298NEGATIVITY BIAS
hedonic episodes, personality traits, etc.) is more nega- has to save to neutralize one act of murder). The cock-
tive than the algebraic sum of the subjective values of roach and the purification rites examples cited at the
those individual entities. The entities being summed beginning of this article embody distinctly the
algebraically are not stimuli, but evaluations; hence, diachronic subtype of negativity dominance.
negativity dominance occurs after we take into any We believe that instances of negativity dominance
possible effect of negative potency and is, in principle, afford us more dramatic and effectual means of
independent of it. Negativity dominance does not re- showing that the core of the positive–negative asym-
quire, operationally, the use of objectively equated or metry in both physical and moral domains may be the
objectively measured stimuli. We consider negativity fact that the corrupting or “devaluing” power of bad
dominance the most robust and most common exem- things is greater than the redeeming power of good
plification of negativity bias, and this article focuses things (Royzman & Kumar, 2001). The image that
primarily on this principle. All of the examples offered comes to mind from the physical domain is that of a
at the beginning of this article illustrate negativity single cancerous growth or germ that radiates itself
dominance. In the purest condition, negativity domi- through and ultimately consumes a perfectly healthy
nance holds that the combination of events of equal but body. The image that comes to mind from the moral
opposite subjective valence will be negative. Thus, if domain is that of a single vice corrupting and pervert-
losing $100 is worse than winning $100 is good, we ing and bringing the moral downfall of an otherwise
have an instance of potency. But if we then find that perfectly good person.
losing $100 is as bad as winning $150 is good, and that
losing $100 and winning $150 is negative, then we
have negativity dominance. Greater Negative Differentiation
Kanouse and Hanson (1972) recognized the partic-
ular importance of negativity dominance. They framed Negativity bias manifests itself in the fact that nega-
the power of negative properties in terms of their abil- tive stimuli are generally construed as more elaborate
ity to interfere with enjoyment of positive aspects, as and differentiated than the corresponding positive
when a rancid taste completely ruins the good taste of a stimuli. This phenomenon of greater negative differen-
soup. They suggested that “negative components of a tiation represents, in our view, yet another facet of the
complex object are overweighted only when the good general negativity bias principle. The most reliable
and the bad are found together in one object, when they finding consistent with this phenomenon is that the vo-
are inseparable” (p. 58). Moreover, as one of us noted cabulary used to describe the qualities of evaluatively
elsewhere (Royzman, 2000), it is precisely when nega- negative phenomena is far richer and more varied than
tive and positive stimuli are “blended together” to form that employed to depict those associated with
a novel gestalt that one sees one of the most remarkable evaluatively positive stimuli (Peeters, 1971), suggest-
manifestations of the dominance principle—“negative ing that our cognition is perhaps more complex, elabo-
overassimilation,” in which a property that is evaluated rated, and fine-tuned when it comes to the occurrences
negatively in its own right “may be judged even more of the former (e.g., Czapinski, 1985). Negative differ-
negatively when … lodged in a positive subject” entiation is distinguished by Peeters and his colleagues
(Rokeach & Rothman, 1965, p.130), so that “irrespon- from the other negativity bias effects under the term in-
sible father” could be judged more negatively than “ir- formational negativity effect, in contrast to the affec-
responsible” (Rokeach & Rothman, 1965) and “loyal tive negativity effect, which includes what we call
martinet” more negatively than “martinet” per se. potency and dominance. Another example of greater
We find it useful to draw a further distinction be- negative differentiation, described later, is the gener-
tween the synchronic (simultaneous) and the ally greater number of negative than positive emotions.
diachronic (successive) manifestations of negativity
dominance. The former concerns the appraisal of nega-
tive and positive components as co-occurring constitu- A Possible Additional Aspect of
ents of a single whole (as in the case of forming a Negativity Bias: Less Adaptation
holistic impression of a person on the basis of a list of to Negative Events
adjectives that describe that person’s negative and pos-
itive traits). Under these conditions, the negative com- There are suggestions in the literature, from studies
ponent would be disproportionately more influential in of pain and stress (see, e.g., Taylor, 1991) and adjust-
determining the overall appraisal than the positive ment to major good and bad life-events (Brickman,
components of comparable magnitude. The diachronic Coates, & Janoff-Bulman, 1978), that there is less adap-
subtype, on the other hand, is revealed in the cancella- tation to negative than positive situations. We will not
tion of positive by negative events, and vice versa (as consider this possibility further for two reasons: (a)
in the case of determining how many lives a murderer there is not much evidence to support this claim; and (b)
299ROZIN & ROYZMAN
even if there was evidence, it would be subject to the in- lated evidence, made excellent use of this strategy in
terpretation that it was a byproduct of negative potency. their article, illustrating extensively how the absence
Careful matching of negative and positive inputs would of the negative matters more than the presence of the
have to be carried out to establish less adaptation to neg- positive across a variety of domains, including health,
ative events as independent of negative potency. parenting, and relationships. Seventh, less logically
sound but often convincing, are findings of a large dis-
parity in effect between a negative and a positive event,
A Note on Method as when there is obviously nothing to match the po-
tency on the positive side of a cockroach touching
The logic of argument for negativity bias is com- food. All seven of these lines of evidence can be sup-
plex, largely because of the difficulty of equating nega- ported by experiments, careful observations, and/or
tive and positive events. At one level, one can compare anecdotal reports.
subjective reactions to objectively equal negative and
positive events; this depends on some sort of accept-
able scale for events, such as money (losing vs. win- Evidence for Negative Bias
ning $100) or temperature. A second approach is to in Different Domains
compare combinations of stimuli equated for subjec-
tive intensity, or more generally, showing that the out- This article is organized in terms of the domains in
come evaluation of combined stimuli is more negative which negativity bias may occur: sensory, memory,
than an algebraic sum of subjective (or objective) in- contamination, impressions of persons, moral judg-
tensities. A third possibility, depending on the claim, is ments, and so forth. Superimposed on these domains is
to show an interaction effect, such that, for example, our organization of the ways in which negativity bias is
mixed negative and positive stimuli become more neg- manifested. This taxonomy can be fit within the Peeters
ative the closer one is to them. A fourth way of making affective and informational distinction. Some of the ar-
a meaningful comparison is to show that a negative eas that we review have received thorough recent re-
event pushes some output into negativity, but a posi- views, in the context of a negative bias, particularly by
tive event that corresponds in some way to it has no ef- Peeters and his colleagues (Lewick, Czapinski, &
fect in the positive direction. This avoids the scaling Peeters, 1992; Peeters, 1971, 1989; Peeters &
problem. A fifth strategy is to show that there is no ex- Czapinski, 1990), Taylor (1991), Cacioppo and his col-
act positive equivalent of a given negative construct. leagues (Cacioppo & Bernston, 1994; Cacioppo,
For example, Baumeister et al. (in press) argued that Gardner, & Bernston, 1997, 1999), and Baumeister and
there is no apparent positive equivalent to the event of his colleagues (Baumeister et al., in press). In those do-
psychic trauma and the associated condition of Post mains, our review will be less thorough and will cite
Traumatic Stress Disorder. A sixth approach is to show principally the conclusion of prior reviews.
that, given two logically related but oppositely
valenced constructs (e.g., pessimism vs. optimism, bad
parenting vs. good parenting), it is the absence or pres- Physiological Arousal
ence of the negative construct that is the principal de-
terminant of an outcome of interest (e.g., recovery), Taylor (1991) recently reviewed this literature and
with the positive counterpart making little or no mar- concluded that there is generally more physiological
ginal contribution. An example of this strategy at work arousal to negative events and that arousal alone is
is afforded by a recent study by Schulz, Bookwala, generally interpreted negatively. She also pointed out
Knapp, Scheier, and Williamson (1996). Drawing on that negative stressors (changes) seem to have more of
the prior research, indicating that pessimism and opti- an effect on health than “equivalent” positive “stress-
mism are better viewed as separate factors rather than ors.” As she admitted, the evidence on all of these
bipolar opposites, these authors examined separately points is not convincing, because there have been few
the effects of dispositional pessimism and direct comparisons, and because it is difficult to equate
dispositional optimism (assessed as responses to four positive and negative events for purposes of compari-
negatively phrased and four positively phrased son. All of the evidence on physiological arousal is di-
subscales of Scheier & Carver’s, 1985, Life Orienta- rected at the demonstration of negative potency (as
tion Test) on survival among 238 patients with ad- opposed to negativity dominance, which involves
vanced cancer. Pessimism was a significant inverse combinations of negative and positive events). How-
predictor of survival at the 8th month’s follow-up for ever, Cacioppo et al. (1999) reviewed evidence from
the younger age group (30–59). On the other hand, op- evoked potentials in humans suggesting disproportion-
timism was not a significant predictor of survival at ately negative outputs from combinations of negative
any age. Baumeister et al., who also cited this and re- and positive inputs (negativity dominance).
300NEGATIVITY BIAS
Sensation and Perception porting on the state of the environment. Pain is the
principal nociceptive system, but others include the
As Schopenhauer (1844/1995) noted more than 100 sensations resulting from empty lungs, a full bladder,
years ago, the absence of pain, unlike pain, has no dis- or certain types of gastrointestinal upset (producing
tinct phenomenological presence: nausea). In each of these cases, there is no obvious
beneceptive input that results from normal function.
We feel pain, but not painlessness. … We feel the de- Troland identified the erotic system as the basic posi-
sire as we feel hunger and thirst; but as soon as it has tive system. Some systems have nociceptive and
been satisfied, it is like the mouthful of food which has beneceptive components, such as taste and smell. In
been taken, and which ceases to exist for our feelings taste, there are more negative (bitter, sour) than posi-
the moment it is swallowed. (p. 575)
tive (sweet) subsystems. Troland noted that decrease in
rate of increase of activity in a nociceptive system, or
Schopenhauer’s (1844/1995) claim is correct for the more clearly, cessation of its activity, may lead to posi-
body interior. With the exception of positive sensations tive sensations. Examples he offered include the plea-
arising in muscles (as in massage), the inside of the sure of release of discomfort from emptying the
body is basically a source of evaluatively negative in- bladder or bowels, or from breathing following un-
put. No news is good news, from the point of view of pleasant symptoms resulting from oxygen deprivation
the body interior. The sensations that arise from the from the lungs. Indeed, hunger and thirst can be
body interior are essentially painful indications that all thought of as nociceptive systems; the reduction of ei-
is not well (Rozin, 1999; Troland, 1928). Thus, from ther surely produces positive affect, but the systems are
the inside point of view, the “positive” state of normal basically negative: we only appreciate satiation as a
function is the essentially neutral default. Not only does positive sensation at the time that it directly follows a
one not go to the doctor when ones’ organs feel good, period of hunger or thirst.
but one does not even notice it. The response to stimuli leading to negative evalua-
However, the body surface, and especially its aper- tions are generally more distinct and intense than those
tures, represent both pleasure and pain (Rozin, 1999; leading to positive evaluations. This is clear in compar-
Troland, 1928). But even on the body surface, there is a ing negative and positive facial expressions to tastants
wider distribution of pain. Pain can be produced any- in both humans (Steiner, 1979) and rats (Grill &
where on the body surface, whereas the loci for pleas- Norgren, 1978).
ant sensations are far more circumscribed, even named Although there is substantial adaptation to contin-
for one set of pleasant sensations: the erogenous ued stimulation in most sensory systems, the pain sys-
zones—there are no labels for “torturogenous zones.” tem stands as a notable exception. Pain, as an indicator
The “we feel pain, but not painlessness” dictum finds of something awry, remains an attention-getting input.
its reformulation in Scitovsky’s (1974) idea of the Furthermore, there is evidence that people adapt more
phenomenological paleness of comforts, goods that are to fragrant than pungent smells (Cometto-Muniz &
dedicated to preserving a minimal level of painlessness Cain, 1992).
and that, so to speak, keep pain at bay (e.g., air condition-
ers). Scitovsky pointed out that we gain little from com-
forts, because we adapt to them quickly (see also Attention and Salience
Frederick & Loewenstein, 1999). People generally
don’t get pleasure from their air conditioning, but would Generally, negative information seems to com-
experience immediate discomfort if it ceased to operate. mand more attention. The intellectual issue raised by
Negative potency (higher psychological intensity of this small literature is the extent to which negative
negative as opposed to positive events) in sensory sys- bias in salience is completely a byproduct of negative
tems is also exemplified and organized by Troland’s potency (greater psychological impact of negative
(1928) analysis of sensory systems into nociceptive events than equivalent positive events), or whether
(indicating harm), beneceptive (indicating benefit), there are some special negative bias features that op-
and neutroceptive (informational, but innately erate particularly in the domain of attention. The last
affectively neutral). The principal neutroceptive sys- study we review (Pratto & John, 1991) controls for
tems are vision and audition, but proprioception and potency and still reveals a negative bias effect. The
pressure and touch are also included. In all of these small number of articles that address negative bias in
systems, it is generally true that high levels of stimula- attention focus on the greater ease of identifying neg-
tion are aversive, but lower levels are often neutral. ative stimuli, the higher speed of locating negative
Troland suggested that the nociceptive and stimuli in a search task, or the greater masking power
beneceptive systems are reporting on the state of the of negative stimuli. The stimuli involved are faces,
organism, whereas the neutroceptive systems are re- words, and social information.
301ROZIN & ROYZMAN
Masking. Ohnesorg (1999) demonstrated that Learning
negative words are more effective as backward masks
than positive words. The results also suggested that on In the domain of learning, we have the opportunity to
repetition as masks, the attention capturing capacity introduce findings from the animal as well as the human
drops faster for positive as opposed to negative words. literature. The basic claim is the existence of negative
potency and is that negative events, serving as reinforc-
ers, produce learning that is more rapid and more resis-
Identification. Steiner (1979) studied the judg- tant to extinction than learning based on comparable
ments by adults of the emotions expressed by infants, positive reinforcers. The latter claim, about resistance to
presented on videotape, while the infant was experienc- extinction, has not been tested, to our knowledge. These
ing negative (sour or bitter) and positive (sweet) tastes. predictions amount to the claim that learning about neg-
He reported that adults are better at judging the negative ative USs is “prepared,” in the sense defined by
faces. Of course, this suggestive study does not distin- Seligman (1970). A third claim is that it should be easier
guish between the salience of the expressions in the in- to reverse innate preferences than innate aversions. This
fants or the recognition ability of the participants. Simi- entire analysis is subject to the problems raised about
larly, and with the same proviso, H. J. Grill (personal comparing negative and positive in the previous section
communication, 1990) reported a greater sense of “ur- on methodological issues. The problem is particularly
gency” in the Norway rat’s response to negative taste as strong in this domain because much of the data come
opposed to positive taste stimuli. from animal research, where there is no accepted way to
equate subjective intensity.
Students of animal learning are generally aware that
Search. There is one thorough and excellent learning with negative events (e.g., escape in the oper-
study in this area, dealing with the search for negative ant framework) is more rapid than learning with posi-
or positive faces in a field (crowd) of other faces tive reinforcers. Perhaps the most striking case is
(Hansen & Hansen, 1988). The authors reported an traumatic avoidance learning, which occurs in a single
“anger superiority effect.” The task is to scan a “crowd” trial, motivated by a single strong electric shock (Solo-
of black-and-white faces of the same person and iden- mon & Wynne, 1954). It seems to be generally true, al-
tify the one face that is discrepant from the others. The though there is no rigorous test nor systematic
discrepant face can be happy or angry, in a background comparison in the literature, that escape and punish-
of neutral or opposite valence (angry or happy, respec- ment are more effective in producing acquisition and
tively) faces. Reaction times are much faster for angry resistance to extinction than their positive equivalents,
than for happy face targets. Further analysis demon- and negative contrast effects may be stronger, on aver-
strates the striking fact that identification of anger faces age, than positive contrast effects.
occurs at about the same time for crowds of four or nine
faces, whereas the search for a happy face takes longer
with the larger crowd. These results suggest that there is
a parallel search for the angry face, a “pop-out” effect, Taste aversions in animals. Conditioned taste
but not for the happy face. The authors proposed that aversions, in both animals (reviewed in Garcia,
there is a possible preattentive parallel search for sig- Hankins, & Rusiniak, 1974; Rozin & Kalat, 1971) and
nals of direct threat. humans (Garb & Stunkard, 1974; Logue, Ophir, &
Pratto & John (1991) measured reaction times for Strauss, 1981), typically occur in a single trial. Indeed,
naming the colors of words in a Stroop test and found conditioned taste aversions are acquired so rapidly, and
that times were longer for undesirable than desirable with such a robust effect, that it has been necessary to
trait words, suggesting an attention-grabbing power reduce the magnitude of the US (often by reducing the
for negative social information. This effect did not ap- potency of the nausea producing procedure) to demon-
pear for negative versus positive nontrait words, oc- strate a learning curve. Positive learning in the food do-
curred when the diagnostic base rate for the negative main rarely occurs with such rapidity; Sclafani &
and positive trait words was controlled, and occurred Nissenbaum (1988) demonstrated the most rapid learn-
when the trait words were balanced for extremity ing, sometimes in a single trial, using a particular type
(eliminating a simple potency interpretation). The au- of carbohydrate (polycose), and also fat reinforcers.
thors reasoned that if attention is selectively diverted However, overall, there is little doubt in the animal lit-
by negative traits, then more should be learned about erature about the greater speed and robustness of condi-
them in an incidental learning situation. On repeating tioned taste aversions, as opposed to preferences.
the Stroop study, they found that there is greater free Zahorik (1979) attempted a direct comparison and con-
recall of the negative as opposed to positive words af- firmed this relation, although her study was not able to
ter the trial. accomplish a convincing demonstration that the nega-
302NEGATIVITY BIAS
tive and positive reinforcers were equated, in some rea- phobias. (However, there is no evidence that these
sonable sense. strong likes are less resistant to extinction than are pho-
The negative bias hypothesis is strongly supported bias or conditioned taste aversions.)
in studies that attempt to reverse innate preferences or Humans may be unique, among animals, in the de-
aversions. The standard one-trial-effective taste aver- velopment of strong and enduring likes (Rozin, 1982).
sion paradigm uses sugar or saccharine solutions, for These are abundant and include the robust reversal of
which there is an innate preference. On the contrary, innate aversions, as in the development of strong pref-
extended efforts to reverse innate aversions to bitter or erences for chili pepper, coffee, horror movies, tragic
irritant sensations in animals, over many trials and drama, and scary rides at amusement parks
months, have generally failed almost completely (McCauley, 1998; Rozin, 1990). It is possible that this
(Rozin, Gruss, & Berk, 1979; Warren & Pfaffman, perhaps unique human feature is an adaptation to cul-
1959). Exceptions are one study with rats using social ture, which requires adherence to, and preferably a de-
mediation (conspecific consuming an irritant diet), sire for, a whole set of cultural values (Rozin, 1982).
which led to a modest preference for a mildly irritating However, whatever the reason for this, these strong
diet (Galef, 1989); one study showing a very gradual positive attachments develop gradually.
development of a preference for piquant crackers by The opposite of a taste aversion or a phobia would
captive chimpanzees, in a situation of extended social be what we call a fetish, or more generally, a passion.
interaction with humans (Rozin & Kennel, 1983); and Passions exist on a grand scale in humans, especially in
one case of a dog that gradually developed a preference the first world, where there is leisure time that would
for piquant foods, in a social and home environment allow for such activities. Passions, which we will de-
(Rozin & Kennel, 1983). Although humans regularly fine as strong liking for things or activities that have no
and gradually develop preferences for many innately obvious biological function, become a major part of
unpalatable foods such as chili pepper, ginger, raw gar- one’s life and a major source of pleasure. Passions in-
lic, coffee, and alcohol, there are practically no cases clude activities such as collecting stamps or bottle or
on record of spontaneous development of such prefer- hub caps, horse or automobile racing, and sports
ences in animals. To the contrary, animals that regu- fanship (Wrzesniewski, Rozin, & Bennett, in press).
larly consume spicy Mexican food (as garbage) in a These passions challenge any straightforward adaptive
rural Mexican setting do not develop a preference for account and are, in a sense, a challenge as well to the
chili pepper, whereas all the surrounding adults over principle of negativity bias. Although there has been
the age of about 5 do (Rozin & Kennel, 1983; Rozin & no experimental research in this area (except for sexual
Schiller, 1980). fetishes, see Rachman & Hodgson, 1968), there is no
reason to believe that passions are rapidly acquired, al-
though they often are highly resistant to extinction.
Taste aversions, phobias, and fetishes or passions
in humans. The usually one-trial, robust, taste-aver-
sion phenomenon has been well documented in hu- Evaluative conditioning. Other than mere expo-
mans, primarily by retrospective questionnaire (Garb sure, evaluative conditioning is the only account for the
& Stunkard, 1974; Logue et al., 1981; Pelchat & Rozin, acquisition of likes and dislikes with both a theoretical
1982), and it appears that nausea is the critical uncondi- basis and abundant supporting laboratory experiments.
tioned stimulus that produces the effect (Pelchat & In evaluative conditioning in humans, a positive or neg-
Rozin, 1982). Conditioned taste aversions have also ative US (e.g., an unpleasant or pleasant picture, face,
been produced in humans under controlled conditions or taste) is contingently paired with a relatively neutral
(e.g., Bernstein, 1978). CS. After a number of trials, the participant’s evalua-
Phobias represent yet another area in which there is tion of the CS moves in the direction of the US (Martin
strong, retrospective evidence for rapid one-trial nega- & Levey, 1978). The initial investigators of this type of
tive learning. Single traumatic incidents with dogs and conditioning, Levey and Martin (1975), concluded that
other animals, in particular, seem sufficient to produce “The effect of negative evaluation was clearly stronger
strong negative responses (Solomon & Wynne, 1954). than that for positive evaluation, and this is consistent
Seligman (1970, 1971) used conditioned taste aver- with our knowledge of aversive conditioning” (p. 224).
sions and phobias as prime examples of what he called Subsequent studies of evaluative conditioning, most
“prepared” learning. However, there is no well-docu- prominently by a group of Belgian investigators led by
mented opposite effect of very rapid and robust posi- Frank Baeyens (e.g., Baeyens, Crombez, Van den
tive learning in humans. Humans come to develop Bergh, & Eelen, 1988), are consistent with this result,
strong likes for many things, including foods, music, although no direct tests of the negative bias hypothesis
and pets. In all of these, so far as we know, the acquisi- have been made. The Baeyens group initially used pos-
tion process is much slower than for taste aversions or itive and negative faces as unconditioned stimuli,
303ROZIN & ROYZMAN
roughly balanced for valence by subject ratings. The re- ative and positive event scheduled to occur on the same
sults supported slightly larger conditioned effects for day, from the vantage point of 1 month ahead versus
the negative stimuli. More recently, the Baeyens group tomorrow. For tomorrow, the event combination was
has shifted to a different paradigm, based on pairing of rated as more negative.
flavors with pleasant or unpleasant tastes (Baeyens, N. E. Miller (1944) offered an account of the
Eelen, Van den Bergh, & Crombez, 1990). This para- steeper negative gradient in terms of the source of mo-
digm was based on a positive finding for pairing of tivation. He pointed out that electric shock is an exter-
sweet tastes with neutral flavors (Zellner, Rozin, Aron, nal event, such that closeness to its source should
& Kulish, 1983). However, the Baeyens group found reasonably increase intensity of response. For food,
the positive pairing to produce only marginal effects, however, there is an underlying motivation of hunger,
whereas pairing with a negative taste produced more which presumably does not vary much as one ap-
robust effects (Baeyens et al., 1990). proaches the goal object. In other words, insofar as
Both conditioned taste aversions and acquired pho- negative motivations are more externalized, closeness
bias fit within the paradigm of evaluative conditioning. should be a more powerful dimension in the negative
Both, as previously mentioned, show a substantial case. Miller also noted that the gradient steepness is
negativity bias. separable from the gradient height and referred to stud-
ies (including the work of Brown) that indicated that
overall strength of motivation affects the height but not
Motivation—Gradient Steepness the slope of the gradient function.
In addition to N. E. Miller’s (1944) external versus
This minimally investigated area provides the only internal motivation account, there are two possible, not
direct evidence for one of the four manifestations of mutually exclusive, accounts of gradient effects. Ac-
negativity bias, the steeper approach gradients for neg- cording to the intensity account, some aspect of the
ative as opposed to positive stimuli. The basic finding preevaluative representation of the relevant stimuli
is that as one approaches a negative entity, in either shows a negative enhancement effect. There is at best
time or space, the aversion for that entity or experience suggestive evidence from taste that intensity of bitter
increases more steeply than the increase in attraction (measured by physiological response) rises more
does for approach to a positive entity (N. E. Miller, steeply than for sweet (Pfaffman, 1960), and it seems to
1944). Note that this relation does not depend as much us quite reasonable that this would also hold true for pain
on careful balancing of the negative and positive enti- or pleasure from the body surface. Because we can pre-
ties as do the potency studies because the finding of in- sume that getting closer in time or space to a stimulus in-
terest is an interaction effect. creases the intensity of its representation, this intensity
The evidence for this potentially basic feature of an- negative bias effect could account for gradient effects.
imal and human motivation is summarized by N. E. However, one could also argue that, given the urgency
Miller (1944). The most convincing study was carried of negative inputs, the input might rise to a maximum
out by Brown (1948). Brown measured the rat’s ten- very rapidly, so that gradient effects might only occur
dency to pull toward a reward (food) at the end of an al- over a small range of time or space. There are sugges-
ley, at different points in the alley; he did the same for tions of a steeper function relating negative entities to
the tendency to pull away from a negative event affective negativity in the contagion literature. The phe-
(shock) at different points in the alley. He reported nomenon of dose insensitivity, documented principally
steeper negative than positive gradients in terms of dis- for negative contagion, indicates that very small doses
tance from site of feeding or shock. This was con- of contact with negative entities (e.g., germs, contact
firmed at two different levels of motivation (amount of with an undesirable other person) produce almost the
food deprivation or intensity of shock). In Brown’s de- maximal effect (e.g., Rozin & Nemeroff, 1990; Rozin,
sign, temporal and spatial closeness are confounded. Markwith, & Nemeroff, 1992).
Although it is likely that the gradient phenomenon The negative gradient effect might also result from
holds for time or space alone, this has not yet been an asymmetry later in the processing system. In the
demonstrated. face of equal recruitment of intensity with closeness, it
Another type of demonstration of the gradient effect is possible that the function relating subjective inten-
would be to show that for events (or roughly simulta- sity to evaluation is what is more steep for negative
neous combinations of events) of mixed positive and than for positive stimuli. We know of no direct evi-
negative characteristics, the net response to these dence on this relation. In either event, it is important to
events becomes more negative the closer one is to them realize, as Cacioppo and his colleagues have indicated
(in space or time). We have recently (Rozin, Kurzer, & (Cacioppo & Bernston, 1994; Cacioppo, Gardner, &
Royzman, 2000) demonstrated this with human partic- Bernston, 1997, 1999), that although the slope is
ipants, who rated their net evaluative response to a neg- steeper for negative events, at low levels, the absolute
304NEGATIVITY BIAS
value of positive affect is higher than is negative affect major effect is reported for delay of recall, such that
(“positivity offset”). longer delays lead to more positive bias. Taylor
(1991), in a review, came to a similar conclusion as
Matlin and Stang and focused her account on compen-
Mood satory responses that minimize negative memories,
which occur gradually over time (accounting for the
Taylor (1991), on reviewing asymmetrical effects increased positivity bias with delay). Thus, on Taylor’s
of negative and positive events, concluded that expec- reasonable view (consistent with Matlin and Stang),
tations of negative events are the strongest determi- the major reason for positivity bias is not that negative
nants of mood. Insofar as negative events have been events are inherently less memorable, but rather that
equated with positive events, she suggested that nega- they are neutralized over time.
tive effects still have a bigger effect on mood. Results on both autobiographical memory and labo-
ratory studies suggest that over modest to long inter-
vals of recall, there is a positivity bias in memory. The
question remains as to whether this bias directly con-
Memory
tradicts the principle of negativity bias. There are two
processes that may be at work in these studies to en-
For fine ideas vanish fast, While all the gross and filthy
last. (W. I. Miller, 1997, p. 70 [Strephon and Chloe vv hance the appearance of a positivity bias. There is
233–234, Poetical Works, 525]) abundant evidence that positive events are much more
frequent than negative events; hence, in an autobio-
The evil that men do lives after them; the good is oft graphical recall study, one would expect more positive
interr’d with their bones. (Shakespeare, quoted in events. This is a substantial bias, but it has been elimi-
Greenblatt, 1997, p. 1565) nated as a total account by a number of studies in which
participant’s recall was checked against diary records
These quotes notwithstanding, the existence of a of positive and negative experiences that they were
negativity bias (negative potency) in memory is con- asked to keep (Holmes, 1970; reviewed in Matlin &
troversial. Baumeister et al. (in press), in their review Stang, 1978).
of the memory literature, concluded in favor of the ex- The second “bias” in recall is the very compensa-
istence of a negative bias in memory, at least for cer- tory processes that are aptly described by Taylor
tain types of information. On the other hand, Taylor (1991). Work is done to reduce the salience of negative
(1991) and Matlin and Stang (1978) argued for greater events in memory. We are inclined to believe that Tay-
potency of positive memories. Our own consideration lor identified the major reason for a positivity bias in
of the literature inclines us to support a positivity bias recall and that the mechanisms she invokes themselves
view in memory. testify to the salience of negative events.
The widespread operation of negativity bias might
be expected to generalize to selective memory for neg-
ative experiences. This prediction might be strength- Contagion
ened by the greater attentional salience of negative
events (reviewed earlier) and abundant evidence for The domain of contagion offers what are perhaps the
deeper processing of negative events (reviewed in most striking instances of negativity bias. It is partly for
Baumeister et al., in press). Although there are some this reason that the striking examples we introduced at
striking examples of predominance of negative bias in the beginning of this article are about contamination.
recall, notably of early childhood memories (Blonskii, The negativity bias in contagion is evidenced in the ter-
1935/1994; Kreitler & Kreitler, 1968), in our view the minology alone: Negative contagion is represented in
preponderance of evidence suggests a positivity bias in the word “contamination” in English. There is no obvi-
this domain. The most thorough treatment is in a chap- ous opposite term for positive contamination (purified
ter reviewing this very issue in a book devoted to dem- has a much more general meaning and does not imply
onstrating a general positivity bias (Polyanna minimal contact, as does contamination).
Principle) by Matlin and Stang (1978). Well over 100 The law of contagion was put forward as one of the
findings, from various paradigms, show a positivity laws of sympathetic magic by three anthropologists
bias in a majority of cases. Matlin and Stang consid- around the turn of the century (Frazer, 1890/1922;
ered three accounts for this: advantages in short-term Mauss, 1902/1972; Tylor, 1871/1974). The basic idea
memory, selective rehearsal, or compensatory pro- is that when entities contact, “essence” passes between
cesses in long-term memory, and find evidence for them and leaves a permanent trace (“once in contact,
each. Most impressive is their multiple regression always in contact”; reviewed in Rozin & Nemeroff,
analysis of 14 variables on degree of positivity bias. A 1990; Nemeroff & Rozin, 2000). Minimal contact is all
305ROZIN & ROYZMAN
that is necessary to allow for the passage of essence. rule is that “any known African ancestry renders one
The early anthropologists saw contagion as a belief of Black” (Lopez, 1996, p. 27). As Lopez pointed out, in
“primitive” peoples, but it is now clear that it is univer- accordance with this rule, “no ‘mixed-race’ applicant
sal (Rozin, Millman, & Nemeroff, 1986). Common ex- was naturalized as White” (p. 27). There exists no his-
amples for Americans include reluctance to consume torical evidence for the positive equivalent of a
foods briefly contacted by worms or cockroaches or to “one-drop” ordinance—that is, a statute whereby one’s
wear clothing that had previously been worn by a dis- membership in a racially privileged class would be as-
liked person. sured by one’s being in possession of “one drop” of the
Negativity bias in the potency domain is intu- racially superior blood (a situation of some bearing
itively clear for Americans and has been demon- here is the determination, often for purposes of mar-
strated in the laboratory (Rozin et al., 1986; Rozin, riage or succession, as to whether a person is of a
Nemeroff, Wane, & Sherrod, 1989). In particular, “royal” line; certainly, in this case, the royal blood can
disliked people produce a larger contagion effect than represent less than half of the total “blood,” but we
do liked people. In addition, as indicated in the intro- know of nothing like a one-drop rule).
duction, in the food domain, there is nothing nearly The Nuremburg laws, promulgated by the Nazis in
as potent on the positive side as a cockroach or an the 1930s to define and persecute Jews, forced a defini-
earthworm on the negative side. tion for Jewishness. The one-drop rule here would
Negativity bias in potency is also illustrated in the have been impossible to enforce; instead, the rule was
Hindu caste system, where contact with lower castes that one Jewish grandparent was sufficient for the des-
produces much more contagion than does contact with ignation as Jewish. It is interesting in this regard that in
higher castes. Among the Hua of New Guinea (Meigs, the affirmative action debate in the United States,
1984), perhaps the most contagion sensitive of all cul- where previously stigmatized and contaminating
tures studied, there is abundant positive contagion, but groups, especially Blacks, are now given preference,
negative contagion remains more powerful. nothing like a one-drop rule has been instituted. It is
There is relatively little data on dominance effects rather general resemblance and associations with
(greater effects of contaminants as opposed to purifi- members of the previously stigmatized groups that
ers, over and above their rated potency). We have some makes one eligible for privileged treatments. Thus,
unpublished data (Rozin & Royzman, 2000) that indi- when “Black was bad,” a one-drop rule justified inclu-
cates that combinations of negative and positive conta- sion in the category, but a much more substantial link
gion (e.g., a sweater worn by a disliked and then a liked is required for Black status now that, in some contexts,
person, or vice versa) is rated more negative than the “Black is good.”
algebraic sum of the subjective evaluative ratings of
each sweater separately.
It is notable that contagion is both a domain of mani- Decision Making
festation of negativity bias and a theory for why it oc-
curs. That is, in general, negative events may have more Loss aversion, one of the most fundamental and
penetrance or contagiousness than positive events. well-documented biases in information processing, is a
quintessential illustration of negativity bias in the form
of potency. The principle of loss aversion, based on the
Asymmetric weighting and racial purity. Cer- prospect function, holds that losses are more negative
tain practices of designating a racial underclass in rela- than corresponding gains are positive (Kahneman &
tion to its ancestral roots proffer another compelling Tversky, 1979; Tversky & Kahneman, 1991). In its
(and politically consequential) example of negativity boldest form, losing $10 is worse than winning $10 is
bias at its bluntest. In the words of Neil Gotanda (cited good. Although we are convinced of the general valid-
in Lopez, 1996), “The metaphor [for defining a racial ity of loss aversion, and the prospect function that de-
underclass] is one of purity and contamination: White scribes and predicts it, we confess that the
is unblemished and pure, so one drop of ancestral Black phenomenon is only realizable in some frameworks. In
blood renders one Black. Black is a contaminant that particular, strict loss and gain of money does not reli-
overwhelms white ancestry” (p. 27). Of particular in- ably demonstrate loss aversion (unpublished data by
terest, from this perspective, is the notorious “one drop the authors). Perhaps the most robust demonstration of
of blood” rule of racial categorization, having its for- loss aversion is in the endowment effect (Kahneman et
mal origins in provisions of the Code Noir (the “Negro al., 1990), in which the loss is framed as loss of a “pos-
Code”) of 1685, designed, in part, to safeguard the “pu- session” and the gain as acquisition of the same posses-
rity” of the White race by eliminating “tainted” blood sion. Under these circumstances, in a number of cases,
(the rule enjoyed considerable vogue in certain parts of the loss of a possession, literally just given at random
postcolonial America, e.g., Alabama, Arkansas). The to the participant, is valued at somewhere around twice
306You can also read
